A novel blind non-decision directed maximum likelihood algorithm for fractionally-spaced nonminimum phase FIR channel identification and equalisation is presented. The algorithm results from using the low signal to noise approximation to the average of the likelihood function with respect to the transmitted data sequence. The channel estimation equation is derived in a closed form. The resulting algorithm has two distinct advantages. The first is that the channel estimates are asymptotically consistent, and the second is that the algorithm is computationally efficient since it only requires the calculation of one eignevector. Simulation results are presented to show the performance of the proposed algorithm.